Software Developer & AI Engineer | IIT Bhubaneswar

I am experienced in distributed systems internals, low-latency backend infrastructure, and production GenAI/Agentic RAG orchestration. With a foundation in competitive programming and core computer science fundamentals, I design and build resilient, high-throughput software architectures from first principles.

• 🎓 Education: Integrated B.Tech from IIT Bhubaneswar
• 🏆 Achievements: Bronze Medalist at Inter-IIT Tech Meet 10.0 (podium across 23 IITs)
• 💡 Competitive Programming: 1,500+ DSA problems solved | LeetCode Peak: 1934 (Top 4%) | CodeChef: 1682 (3-Star) | Codeforces: 1345
• 🌐 Website: subhransu.dev

• Languages: Java, C++, Python, JavaScript, TypeScript, SQL
• Core CS: Docker, Kubernetes basics, Git/GitHub, Linux, OS, Computer Networks, DBMS, System Design (HLD/LLD)
• Backend & Systems: Node.js, Express.js, FastAPI, RabbitMQ, Microservices, RESTful APIs, Concurrency & Multithreading
• Distributed Systems & Storage: Amazon Dynamo Model (AP Systems), Quorum Consensus, Consistent Hashing, Vector Clocks, Merkle Trees, Gossip Protocol
• Databases & Data Engineering: PostgreSQL, MongoDB, Redis, RocksDB, Google BigQuery, Qdrant Vector DB, Looker Studio
• AI & GenAI Orchestration: LangGraph, Agentic RAG, Self-RAG, Corrective RAG, AST-guarded Text-to-SQL, PyTorch, Hugging Face

Java implementation of the seminal Amazon Dynamo distributed storage paper.

• Built an AP distributed store with asynchronous quorum consensus ($N=3, W=2, R=2$), achieving 12 ms p99 write latency.
• Implemented consistent hashing with MD5 virtual nodes (sub-4% key distribution deviation), vector clocks for causally concurrent updates, and Merkle tree anti-entropy for background replica sync.

Stateful Agentic RAG & Secure Text-to-SQL Execution Engine.

• Architected a multi-path reasoning workflow using LangGraph with self-correcting query rewrite loops and Qdrant hybrid retrieval (Dense + Sparse RRF).
• Integrated an AST-guarded Text-to-SQL engine with 9-layer defensive guardrails to validate queries prior to read-only database execution.

Decoupled real-time food delivery architecture.

• Engineered 6 autonomous Node.js/TypeScript microservices communicating asynchronously over RabbitMQ message queues.
• Built real-time driver dispatch using Socket.IO and MongoDB 2dsphere geospatial indexing, with idempotency keys and HMAC validation for transactions.

• Data Engineering Intern @ HealthyDay
  • Engineered Google BigQuery infrastructure supporting 600K+ registered users.
  • Optimized 150+ SQL queries/pipelines and designed 50+ reusable analytical views powering real-time Looker Studio retention and operations dashboards.
• AI Engineer Intern @ Hanyaa Auto Technologies
  • Fine-tuned Dia-TTS (1.6B parameters) across English, Hindi, and Telugu using custom datasets and dynamic LR schedules.
  • Shipped an automated text-to-animated-video narration pipeline with chunking and seed-locked speaker consistency.

• Portfolio: subhransu.dev
• Email: subhransu.nayak.connect@gmail.com
• LinkedIn: linkedin.com/in/subhransu-p-nayak
• Resume: View / Download